visual analisys of cross correlation and correlaton matrices

For each model:

  1. Choose model
  2. Choose style
  3. matrix_crosscorrelation all (all epochs): heatmap + histogram of log(eigenvalues) + eigenvalues scatterplot
  4. matrix_crosscorrelation small (first 500 epochs): heatmap + histogram of log(eigenvalues) + eigenvalues scatterplot
  5. matrix_crosscorrelation vsmall (first 100 epochs): heatmap + histogram of log(eigenvalues) + eigenvalues scatterplot
  6. Comparission of eigenvalues for different samples of the same model

get_crosscorrelation(paths2, k=1, max_epoch=400, folder='matrix_crosscorrelation_small/') get_crosscorrelation(paths3, k=1, max_epoch=400, folder='matrix_crosscorrelation_small/') get_crosscorrelation(paths4, k=1, max_epoch=400, folder='matrix_crosscorrelation_small/')

Show all matrices

Eigenvalues spectrum all

pwd = 'model_50_50/model_50_50' folder = 'matrix_crosscorrelation/' d=3 show_eigenvalues_all(folder, pwd, d, n_of_samples, cmap, vmin_corr, vmax_corr, mineigenval, maxeigenval, cutoff = 1e-15)

pwd = 'model_50_50_50/model_50_50_50' folder = 'matrix_crosscorrelation/' d=4 show_eigenvalues_all(folder, pwd, d, n_of_samples, cmap, vmin_corr, vmax_corr, mineigenval, maxeigenval, cutoff = 1e-15)

pwd = 'model_50_50_50_50/model_50_50_50_50' folder = 'matrix_crosscorrelation/' d=5 show_eigenvalues_all(folder, pwd, d, n_of_samples, cmap, vmin_corr, vmax_corr, mineigenval, maxeigenval, cutoff = 1e-15)

Show eigenvalues histogram

Plots one by one

VSmall size: 500 epochs